Hunting for reliable low latency, HFTs look to novel techs in 2023

WatersTechnology looks at advancements in market data latency technology and what role the cloud can play.

The need for speed was alive and well in 2023, as trading firms, exchange operators, and infrastructure providers continued their quests to shave off milliseconds and gain an edge. 

Over the past year, WatersTechnology has reported on developments in the data latency space, including the use of shortwave radio frequencies, optimized fiber and microwave bandwidths, and whether or not cloud fits into the picture. 

Bouncing signals off the ionosphere

High-frequency trading (HFT) firms typically use a combination of fiber and microwave connectivity, as well as hardware such as field-programmable gate arrays (FPGAs), to receive and process market data or send trade order messages with ultra-low latency. 

But there’s a new—or rather, old—technology that could complement or change their low-latency strategies. Shortwave radio frequencies, or high-frequency broadcasting—which uses the 3 to 30 MHz frequency band—has been around since the early 1900s, but has only recently become a realistic option for trading firms. 

There is already tangible evidence that shortwave technology is used for trading, as Stefan Schlamp, head of quantitative analytics at Deutsche Börse, explained in a LinkedIn post. The best use of shortwave technology could be to look at instruments that are tightly correlated, widely traded, and traded across different geographic markets. Stéphane Tyc, co-founder of low-latency microwave bandwidth provider McKay Brothers, told WatersTechnology that some important signals could be movements in the S&P 500, some futures on oil, bond futures or agricultural futures instruments. 

As with any technology, shortwave has its drawbacks, such as bandwidth constraints, availability and reliability limitations, vulnerability to atmospheric conditions, and lack of regulatory clarity. 

That aside, some firms are providing shortwave service to capital markets firms. One example is Shortwave Solutions, a transoceanic infrastructure connectivity provider spearheaded by Alex Pilosov, who was the first to build a microwave link from New York to Chicago. 

Another firm is Tel Aviv-based Raft Technologies, a shortwave service provider founded in 2013, which in February operated six ultra-low latency links between the US, Europe, and Asia. (Click here for a deeper dive into these companies’ offerings.)

Let’s benchmark it 

Interestingly, Raft later engaged the Securities Technology Analysis Center (Stac) to audit the latency and throughput of two of its HF links from Chicago to London—Raft HF Ultra Link and Raft HF Super Link. 

Stac’s report, which is published on its website, is the first to document the performance of an independently audited HF link for sending cross-exchange signals. 

Stac president Peter Nabicht told WatersTechnology that Stac devised a series of latency and throughput tests for HF radio with input from financial firms, which emulate multiple usage patterns of trading firms. 

Stac tested the data originating in the CyrusOne Aurora datacenter facility (formerly CME’s own co-location facility, where it still hosts its market technology) and terminating in Equinix’s LD4 London facility. 

According to the report, during the hours covered by its service-level agreement (SLA), from 12 pm UTC to 11:59 pm UTC, the mean latency of Raft’s HF Ultra Link from Chicago to London was 24.8 milliseconds. The daily adjusted “goodput”—which measures how fast and accurately useful data traveled from one point to another—was between 77% and 94%, and the link’s uptime was 98%, while usable time was 93%. Comparatively, for Raft’s HF Super Link, during SLA hours, the mean latency was 27.7 milliseconds. The daily adjusted goodput was between 92% and 96%, and the link’s uptime was 98%, while usable time was 94%. 

The report stresses that HF radio links are not deterministic. This is unlike fiber—and to a great extent, microwave as well—said Lior Katz, vice president of marketing and business development at Raft. “This means that what you’re sending, you’re not necessarily getting 100% of that on the other end. That’s why there’s an emphasis on that in the report; getting a detailed performance breakdown is useful for appreciating the concept of an undeterministic data link,” he said. 

While trading firms typically like determinism, they also like the lowest latency they can find. This was why it was important for Stac to quantify the latencies, goodput and the ranges, and provide that information to financial firms. 

“If they like the latency enough, they will figure out ways to incorporate it [in ways] that maybe don’t require as much determinism as they’re used to in other communication mediums. And in doing that, they just want to understand the tradeoffs they’re in and the ranges of what they find acceptable or not—and that’s where having this data is really important,” Nabicht said. 

The use case for shortwave may extend beyond HFT firms, too. “I think everybody uses a different definition of HFT,” Nabicht says. “I think the important thing is they’re primarily serving traders with low-latency needs. Those trading firms often participate in high-frequency trading, but that doesn’t mean they exclusively trade with high frequency.” 

Old-school fiber and wireless 

Developments in shortwave technology don’t mean a decrease in fiber and microwave services; rather, connectivity over fiber and wireless services continues to grow. 

One such example is Intercontinental Exchange’s work to roll out a transatlantic data service offering ultra-low latency data between the US and Europe. It combines the current wireless services of the ICE Global Network with low-latency fiber for a market data offering between Aurora in Illinois, and Europe. 

Nicolas Bonnet, director of global connectivity product at ICE, said the company is already moving data from Aurora to New Jersey and between Aurora and Asia. So, it was looking to bring that data into Europe as well. 

“We’ve managed to combine the relevant assets to make this into an offering and we can now go live,” he said. 

The route starts in Illinois and runs to New Jersey, before traveling across the Atlantic via an underwater cable system to London, where the signal fans out. 

ICE planned to launch the service in October, with the first route going eastbound to datacenters in Slough and Basildon in the London area (The Basildon facility, originally built by NYSE as its primary datacenter in Europe, is the home of ICE Data Services in Europe). 

The launch also includes access to Frankfurt, Germany, the home of Deutsche Börse and Eurex. In November, access extended to the London Stock Exchange’s Telehouse North Two datacenter in Docklands, London, and the newly relocated Euronext datacenter in Bergamo, Italy. 

All five endpoints allow access to significant derivatives, equity, fixed-income and futures markets in Europe, with London serving as the gateway to the rest of Europe. Target users of this offering range from HFT firms to prop shops and tier-one banks. 

Bonnet added that this rollout will help ICE’s future expansion plans. “Stage one is about moving data from Aurora to multiple locations in Europe,” he said. “It’s just stage one because we are looking at the bigger picture,” and pushing the connection farther eastward. 

The second phase could include another data source from the US delivered to Europe, or it could be a new data source from Europe delivered to the US using the same combination of technology. 

Changing delivery models 

It’s noteworthy that ICE has publicly declared that it has no plans to outsource any critical infrastructure to the cloud. This stands in stark contrast to the substantial cloud partnerships that Nasdaq, CME Group, and the London Stock Exchange Group have forged with Amazon, Google, and Microsoft, respectively.

“We really want to control that technology, and we actually think we do a very good job of that,” said ICE’s futures exchanges senior vice president Trabue Bland, speaking at the FIA Boca conference in March. “I can’t even imagine the thought of outsourcing that critical infrastructure to the cloud.” 

Bland added that ICE CEO Jeff Sprecher and CTO Mayur Kapani were confident the exchange was better placed than large tech providers to manage data storage projects. “If you asked either one of them or any of the management team if they felt they could do something better than AWS, they’ll say, ‘Yes,’” Bland said. 

But as exchanges such as Nasdaq successfully move market infrastructures to the cloud, ICE is no doubt watching closely, and may end up revisiting that decision.

Exchanges have long used IP multicast to distribute data to market participants. Until recently, it hasn’t been possible to replicate multicast in cloud environments, making it unsuitable for functions such as matching engines and low-latency data distribution—a barrier to adoption in some areas. 

“Any function that has a timeframe of one millisecond to one day after a trade, from surveillance to settlement—those areas are easy to move to the cloud,” said Alex Wolcough, CEO of UK-based consultancy GreenBirch. “Moving all the non-matching engine areas into the cloud will probably happen, but maybe matching engines will prove too hard,” especially for those who haven’t solved the multicast issue. 

Exchanges that have cloud partnerships must find a way to implement or replicate multicast in the cloud, find another technology to achieve similar results, or keep the more sensitive processes on dedicated hardware. 

While multicast feeds are still an issue for exchanges and cloud providers, it is getting easier to connect to datafeeds in the cloud. “It’s still a challenging issue to address, but AWS has done it, the other players will address it, and it will become a non-issue,” said Stephane Dubois, chief product officer at desktop and feeds provider Quodd. 

Three years ago, AWS built a way to emulate multicast in the cloud within its Transit Gateways—hubs that connect AWS clouds to companies’ on-premises networks. 

“The only difference is how the multicast occurs and the performance of it,” said John Kain, head of worldwide financial services market development at AWS

No longer a pipe dream 

The challenge of multicasting market data in the cloud is what network provider and market data carrier Colt Technology Services, along with AWS, have been working on

In October 2022, the pair completed a cloud co-location proof-of-concept (PoC) that demonstrates the viability of hosting and distributing multicast data in the cloud for capital market customers. 

Wesley Elder, capital market product director at Colt, told WatersTechnology that the purpose behind the PoC was to determine if that was even a possibility. 

“In order to deliver any kind of solution, we first had to understand what the cloud is capable of. Is this just a pipe dream? Is this something that will ever be a reality?” he said. But AWS was confident that once Colt could get the data in, it would be able to support the bandwidth, traffic, and everything else necessary. 

The initial PoC involved Colt and AWS building virtual distribution points of presence in the AWS cloud. This led to Colt launching its multicast data services for capital markets with AWS in June 2023. 

A lot of work went into getting the service up and running. This included solving the issue that cloud providers won’t accept multicast data, in case it floods and overutilizes the network. So, for the service, Colt converts the data into a unicast-type packet, allowing the data to get through the AWS infrastructure. Then, Colt converts it back to multicast to distribute—a process that takes “a few microseconds.” 

But for the latency-sensitive trading desks, this process would add a few microseconds to the time it takes to receive market data. While HFT firms will most likely not use this service for trading purposes, Elder said there are other instances that they might benefit from the service, such as using it to test entry into a new market instead of building out co-location, as getting racks and hardware for connectivity can be challenging and costly. 

While HFT firms tend to be extremely latency-sensitive, certain workloads may be less so. 

“[These might include] back testing or perhaps looking at new markets they don’t have today. Rather than going to, say, Thailand, and building a whole thing, we could get that data to them, they can do some testing, and then if it makes sense, then they go and build co-location,” Elder said. “So it’s more of a test bed or a sandbox for them to test and then think about moving into new markets.” 

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact or view our subscription options here:

You are currently unable to copy this content. Please contact to find out more.

Managing the FX challenge for T+1

As firms prepare for T+1 in May 2024, DTCC’s Val Wotton says they should also consider the complexities for cross-border trades. 

You need to sign in to use this feature. If you don’t have a WatersTechnology account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here